TWI555342B - Method of removing deblocking artifacts - Google Patents

Description

Clearing the method of deblocking human noise

The present invention relates to a deblocking filtering method for erasing blocky artifacts in image reconstruction, and more particularly to a method of determining boundary strength at transform edges and respective edges of prediction boundaries.

In order to compress video material, various video formats have been developed. For example, these video formats have MPEG-2, MPEG-4, and H.264/MPEG-4 AVC. As a successor to H.264/MPEG-4 AVC, high efficiency video coding (HEVC) is currently being developed in conjunction with ISO/IEC MPEG and ITU-TVCEG.

According to HEVC, a picture-by-picture is divided into a plurality of maximum coding units (LCUs), and one or more coding units of each LCU are coded by using prediction blocks generated by inter-frame prediction or intra-frame prediction. A difference between the initial block and the predicted block is transformed to generate a transform block, and the transform block is quantized using one of a plurality of predetermined quantization matrices using a quantization parameter. The quantized coefficients of the quantized block are scanned by a predetermined scan type and then entropy encoded. The inverse quantization and inverse transform quantize the coefficients to produce a residual block, and the residual block is combined with the prediction block to produce a reconstructed image. The deblocking filter is used to adaptively filter the reconstructed image to remove block artifacts.

Figure 1 is a schematic diagram of a process for determining boundary strength according to HEVC being developed.

As shown in Figure 1, the boundary is first determined on each edge of all 4x4 blocks. Intensity, and then the final boundary strength is determined at the edge of the 8x8 block. The edge of this 8x8 block consists of the edges of two consecutive 4x4 blocks. The boundary strength of the edges of this 8x8 block is determined as the maximum boundary strength of the two edges constituting the edge of the 8x8 block. However, since the deblocking filtering technique described in HEVC under development is too complicated, this technique deteriorates coding and decoding performance.

Therefore, the specifications of HEVC under development are focused on techniques that reduce the complexity of deblocking filtering and improve the performance of deblocking filtering. In particular, since HEVC concentrates on encoding and decoding ultra high definition images, the new technology can reduce the complexity of operations and can reduce the memory capacity required to perform deblocking filtering.

The present invention relates to a method of deblocking filtering that reduces the amount of data stored in memory and effectively reduces the complexity of the operation of clearing block human noise.

An aspect of the present invention provides a method for clearing block human noise, comprising: determining a boundary strength of each 4 sample edge located in an 8×8 sample mesh, and if the boundary strength is not equal to zero, determining whether deblocking filtering is Executed at the 4 sample edge, if deblocking filtering is performed at the 4 sample edge, the deblocking filter is selected and the 4 sample edges are filtered with the selected filter.

The method according to the invention determines the boundary strength of each of the 4 sample edges located in the 8x8 sample grid, and if the boundary strength is not equal to zero, determines if the deblocking filter is at 4 samples Edge execution, if deblocking filtering is performed at 4 sample edges, the deblocking filter is selected and the 4 sample edges are filtered with the selected filter. Therefore, the complexity of the calculation required to determine the boundary strength according to the present invention is reduced by 50% or more when compared with the HEVC being developed. At the same time, the memory capacity and bandwidth required to determine the boundary strength are reduced by 50% or more without deteriorating the image quality.

1000‧‧‧Dynamic picture coding equipment

1010‧‧‧Screen division unit

1020‧‧‧Transformation unit

1030‧‧‧Quantification unit

1040‧‧‧ scan unit

1050‧‧‧ Entropy coding unit

1060‧‧‧ In-frame prediction unit

1070‧‧ inter-frame prediction unit

1080‧‧‧ inverse quantization unit

1090‧‧‧ inverse transformation unit

1100‧‧‧ Post Processing Unit

1110‧‧‧ Picture storage unit

1120‧‧‧Subtraction unit

1130‧‧‧Addition unit

2000‧‧‧Dynamic picture decoding device

2010‧‧‧Entropy decoding unit

2020‧‧‧Reverse Scanning Unit

2030‧‧‧ inverse quantization unit

2040‧‧‧ inverse transformation unit

2050‧‧‧ In-frame prediction unit

2060‧‧ inter-frame prediction unit

2070‧‧‧ Post Processing Unit

2080‧‧‧ Picture storage unit

2090‧‧‧Addition unit

LCU‧‧‧Maximum coding unit

P0 ₀ ~p0 ₃ ‧‧‧sample

Q0 ₀ ~q0 ₃ ‧‧‧sample

P, Q‧‧‧

A’, A, B, C, D, D’, G, H, F’F, E‧‧‧ Sports Information

bS‧‧‧Boundary strength

1 is a schematic diagram of a process of determining boundary strength according to HEVC being developed; FIG. 2 is a block diagram of a dynamic picture coding apparatus according to the present invention; and FIG. 3 is a block diagram of a dynamic picture decoding apparatus according to the present invention; Figure 4 is a flow chart of a deblocking filtering process according to the present invention; Figure 5 is a schematic diagram of a method for determining boundary strength according to the present invention; and Figure 6 is a schematic view of a 4-sample edge according to the present invention; Figure 7 is a schematic diagram showing an exemplary configuration of a prediction unit according to the present invention; Figure 8 is a schematic diagram of motion information in a buffer on a storage line according to the present invention; and Figure 9 is for use according to the present invention A schematic diagram of determining the position of a sample whose block edge is filtered.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. however, The invention is not limited to the exemplary embodiments disclosed below, but can be implemented in various ways. Therefore, many other modifications and variations of the present invention are possible, and it is understood that the present invention may be practiced otherwise than as specifically described.

The dynamic picture coding apparatus and dynamic picture decoding apparatus according to the present invention may be user terminals such as a personal computer, a personal mobile terminal, a mobile multimedia player, a smart phone, or a wireless communication terminal. The image encoding device and the image decoding device may include a communication unit for communicating with various devices for storing various programs and storing memory for encoding or decoding images.

Figure 2 is a block diagram of a dynamic picture encoding device 1000 in accordance with the present invention.

Referring to FIG. 2, the dynamic picture coding apparatus 1000 includes a picture division unit 1010, a transformation unit 1020, a quantization unit 1030, a scanning unit 1040, an entropy coding unit 1050, an intra prediction unit 1060, and a frame. The inter prediction unit 1070, an inverse quantization unit 1080, an inverse transform unit 1090, a post processing unit 1100, a picture storage unit 1110, a subtraction unit 1120, and an addition unit 1130.

The picture division unit 1010 divides a picture or a slice into a plurality of maximum coding units (LCUs) and divides each LCU into one or more coding units. The size of the LCU can be 32 x 32, 64 x 64 or 128 x 128. The picture division unit 1010 determines a prediction mode and a division mode of each coding unit.

The LCU contains one or more coding units. The LCU has a recursive quadtree structure to specify the partitioning structure of the LCU. The parameters specifying the maximum size and minimum size of the coding unit are included in the sequence parameter set. The segmentation structure is specified by one or more split coding unit flags. Edit The size of the code unit is 2N x 2N. If the size of the LCU is 64×64 and the size of the minimum coding unit (SCU) is 8×8, the size of the coding unit may be 64×64, 32×32, 16×16 or 8×8.

A coding unit contains one or more prediction units. In intra-frame prediction, the size of the prediction unit is 2N×2N or N×N. In inter-frame prediction, the size of the prediction unit is specified by the split mode. If the coding unit is symmetrically divided, the division mode is one of 2N×2N, 2N×N, N×2N, and N×N. If the coding unit is asymmetrically split, the division mode is one of 2N×nU, 2N×nD, nL×2N, and nR×2N.

A coding unit contains one or more transform units. The transform unit has a recursive quadtree structure for specifying the partition structure of the coding unit. The segmentation structure is specified by one or more split transform unit flags. The parameters specifying the maximum size and minimum size of the transform unit are included in a sequence of parameter sets.

Transform unit 1020 transforms the residual signal to produce a transform block. The residual signal is transformed based on the transform unit. The residual block is obtained by subtracting the prediction block from the initial block, and the prediction block is generated by the intra-frame prediction unit 1060 or the inter-frame prediction unit 1070.

Different transformation matrices are used according to the prediction mode (in-frame prediction mode or inter-frame prediction mode). Moreover, in the intra-frame prediction mode, the transformation matrix can be adaptively determined based on the intra-frame prediction mode. Two one-dimensional transformation matrices (horizontal matrix and vertical matrix) are used to transform the transform unit. For example, in the horizontal intra-frame prediction mode predicted within the frame, since the residual signal may have vertical directivity, a DCT-based integer matrix is used for the vertical direction and a DST-based or KLT-based integer matrix is used for the horizontal direction. In the intra-frame prediction mode predicted within the frame, the DCT-based integer matrix is used for horizontal direction and is based on DST or based KLT's integer matrix is used in the vertical direction. Alternatively, the type of the transformation matrix is determined based on the size of the transform unit.

Quantization unit 1030 determines a quantization parameter used to quantize the transform block. The quantization parameter is a quantization step size. A quantization parameter is determined for each quantization unit. The quantization unit is a coding unit that is greater than or equal to a predetermined size. The predetermined size is the minimum size of the quantization unit. The quantization unit having the smallest size is referred to as a minimum quantization unit. If the size of the coding unit is equal to or larger than the minimum size of the quantization unit, the coding unit becomes a quantization unit. A plurality of coding units may be included in the minimum quantization unit. The minimum quantization unit can be an 8x8 block or a 16x16 block. This minimum size can be determined for each picture.

Quantization unit 1030 generates a quantization parameter predictor and generates a differential quantization parameter by subtracting the quantization parameter predictor from the quantization parameter. The differential quantization parameter is entropy encoded.

The quantization parameter predictor is generated as follows.

[First Embodiment]

The quantization parameters of the left coding unit, the upper coding unit, and the upper left coding unit are successively retrieved in the following order. A quantization parameter predictor is generated using one or two available quantization parameters. For example, the first available quantization parameter is set as a quantization parameter predictor. Alternatively, the average of the first two available quantization parameters is set as a quantization parameter predictor, and if only one quantization parameter is available, the available quantization parameter is set as a quantization parameter predictor.

[Second embodiment]

The current coding unit may have no left coding unit, upper coding unit, and upper left coding unit. In other words, the current coding unit has a pre-coding unit in the coding order. such, The quantization parameters of the adjacent coding unit and the pre-coding unit immediately adjacent to the current coding unit may be used to generate a quantization parameter predictor. The quantization parameters are retrieved in the following order: 1) quantization parameters of the left adjacent coding unit; 2) quantization parameters of the upper adjacent coding unit; 3) quantization parameters of the upper left adjacent coding unit; and 4) quantization parameters of the pre-coding unit.

Alternatively, the quantization parameters are retrieved in the following order: 1) quantization parameters of the left adjacent coding unit; 2) quantization parameters of the upper adjacent coding unit; and 3) quantization parameters of the pre-coding unit.

When two or more quantization parameters are available, the average of the first two available quantization parameters is set as a quantization parameter predictor, and when only one quantization parameter is available, the available quantization parameter is set as a quantization parameter predictor. For example, if the quantization parameters of the left and upper coding units are available, the average of the upper quantization parameters is left set as the quantization parameter predictor. If only one of the quantization parameters of the left and upper coding units is available, the average of the available quantization parameters and the quantization parameters of the pre-coding unit is set as the quantization parameter predictor. If the quantization parameters of the left and upper coding units are not available, the quantization parameter of the pre-coding unit is set as the quantization parameter predictor. The average is rounded off.

Quantization unit 1030 quantizes the transform block using a quantization matrix and quantization parameters to produce a quantized block. A quantized block is supplied to the inverse quantization unit 1080 and the scanning unit 1040.

The scanning unit 1040 vectorizes the block to apply the scan mode to scan the quantized coefficients and transform the quantized coefficients into one-dimensional quantized coefficient components.

In the intra-frame prediction mode, the distribution of the quantized transform coefficients varies depending on the intra prediction mode and the size of the transform unit. Therefore, based on the intra prediction mode and transform unit The size determines the scan mode. The scan mode can be selected in sawtooth scan, vertical scan, and horizontal scan. The sawtooth scan can be replaced by a diagonal scan.

For example, if the size of the transform unit is equal to or smaller than 8×8, the horizontal scan is selected for a predetermined number of adjacent intra-frame prediction modes of the vertical mode and the vertical mode, and is a predetermined number of adjacent frames in the horizontal mode and the horizontal mode. The prediction mode selects a vertical scan and selects a sawtooth scan or a diagonal scan for other in-frame prediction modes. When the size of the transform unit is greater than 8x8, a sawtooth scan or a diagonal scan is selected for all intra-frame prediction modes.

In the inter-frame prediction mode, a predetermined scan mode is used. The predetermined scan mode can be a sawtooth scan or a diagonal scan.

When the size of the transform unit is greater than a predetermined large hour, the quantized coefficients are divided into a plurality of subsets and then scanned. The predetermined size may be 4 x 4. The scan mode used to scan the subset is the same as the scan mode used to scan the quantized coefficients of each subset. The quantized coefficients of each subset are scanned in the reverse direction. The subset is also scanned in the reverse direction.

A parameter indicating a last non-zero position is encoded and sent to the decoder. The last non-zero position specifies the position of the last non-zero quantized coefficient in the transform unit. Parameters indicating the location of the last non-zero quantized coefficient in each subset are also sent to the decoding device.

The inverse quantization unit 1080 inversely quantizes the quantized coefficients. The inverse transform unit 1090 inverse transforms the inverse quantized coefficients to generate a residual signal.

The addition unit 1130 adds the residual signal generated by the inverse transform unit 1090 to the prediction signal generated by the intra-frame prediction unit 1060 or the inter-frame prediction unit 1070. Subtraction unit 1120 subtracts the predicted samples from the initial samples to produce residual signals.

The post-processing unit 1100 performs a deblocking filtering process, a sample adaptive offset process, and an adaptive loop filtering process.

A deblocking filtering process is performed to clear the block artifacts that appear in the reconstructed picture.

After performing the deblocking filtering process, a sample adaptive compensation process is performed to reduce the difference between the initial sample and the reconstructed sample. Determine if each picture or slice map performs a sample adaptive compensation process. The picture or slice map can be divided into a plurality of compensation areas, and the type of compensation for each area can be determined. There are four edge compensation types and two band compensation types. If the compensation type is one of the edge compensation types, the edge type of each sample in the compensation area can be determined, and the compensation corresponding to the edge type is added to each sample. The edge type is determined by comparing the current sample with two adjacent samples.

The adaptive loop filtering process is performed by comparing the reconstructed image with the initial image to obtain a filter coefficient. The filter coefficients are applied to all samples in a 4x4 block or an 8x8 block. It is determined whether adaptive loop filtering is performed for each coding unit. Therefore, the size and coefficient of the loop filter can be changed on the basis of the coding unit.

The picture storage unit 1110 receives the reconstructed picture from the post-processing unit 1100 and stores it in the memory. The picture is a frame-based picture or a field-based picture.

The inter-frame prediction unit 1070 performs motion determination using one or more pictures stored in the picture storage unit 1110, and determines one or more reference picture indices specifying one or more reference pictures and one or more motion vectors. Inter-frame prediction unit 1070 utilizes the one or more reference picture indices and one or more motion vectors to generate a prediction block.

The intra-frame prediction unit 1060 determines an intra-frame prediction mode of the current prediction unit and generates a prediction block using the intra-frame prediction mode.

The entropy encoding unit 1050 entropy encodes the quantized coefficient component received from the scanning unit 1040, the intra-frame prediction information received from the intra-frame prediction unit 1060, and the motion information received from the inter-frame prediction unit 1070.

Figure 3 is a block diagram of a dynamic picture decoding device 2000 in accordance with the present invention.

The dynamic picture decoding device 2000 includes an entropy decoding unit 2010, an inverse scanning unit 2020, an inverse quantization unit 2030, an inverse transform unit 2040, an intra-frame prediction unit 2050, an inter-frame prediction unit 2060, and a post-processing unit. 2070, a picture storage unit 2080 and an addition unit 2090.

The entropy decoding unit 2010 extracts and entropy decoded intra-frame prediction information, inter-frame prediction information, and quantized coefficient components from the received bit stream. The entropy decoding unit 2010 transmits the inter-frame prediction information to the inter-frame prediction unit 2060, transmits the intra-frame prediction information to the intra-frame prediction unit 2050, and transmits the quantized coefficient component to the inverse scanning unit 2020.

The inverse scan unit 2020 transforms the quantized coefficient component into a two-dimensional quantized block using an inverse scan mode.

In the intra-frame prediction mode, the inverse scan mode is selected based on the intra-frame prediction mode and the size of the transform unit. The reverse scan mode can be selected in zigzag scanning, vertical scanning, and horizontal scanning. The sawtooth scan can be replaced by a diagonal scan.

For example, if the size of the transform unit is equal to or smaller than 8×8, the horizontal scan is selected for the predetermined number of adjacent intra-frame prediction modes of the vertical mode and the vertical mode, and is the horizontal mode and A predetermined number of adjacent intra-frame prediction modes of the horizontal mode selects a vertical scan and selects a sawtooth scan or a diagonal scan for other intra-frame prediction modes. When the size of the transform unit is greater than 8x8, a sawtooth scan or a diagonal scan is selected for all intra-frame prediction modes.

In the inter-frame prediction mode, a predetermined scan mode is used. The predetermined scan mode can be a sawtooth scan or a diagonal scan.

If the size of the current transform unit is greater than a predetermined size, the quantized coefficient components are inversely scanned on a subset basis to construct a quantized block. This subset has a predetermined size. The predetermined size can be 4 x 4. If the size of the transform unit is equal to the predetermined size, the quantized coefficient component of the transform unit is inversely scanned to construct a transform unit. When the quantized coefficient components are inversely scanned on a subset basis, the same inverse scan mode is used for the quantized coefficient components of each subset.

Multiple subsets are inversely scanned in the reverse direction. The quantized coefficient component is also inversely scanned in the reverse direction. The inverse scan mode applied to the quantized coefficient components to construct the subset is the same as the inverse scan mode applied to the subset of multiple constructs. The inverse scan unit 2020 performs an inverse scan using parameters indicating the position of the last non-zero quantized coefficient of the transform unit.

The inverse quantization unit 2030 receives the differential quantization parameter from the entropy decoding unit 2010 and generates a quantization parameter predictor to obtain a quantization parameter of the current coding unit.

The quantization parameter predictor is generated as follows.

[First Embodiment]

The quantization parameters of a left coding unit, an upper coding unit, and an upper left coding unit are successively retrieved in the following order. A quantization parameter predictor is generated using one or two available quantization parameters. For example, the first available quantization parameter is set as a quantization parameter predictor. Or the first two The average of the available quantization parameters is set to the quantization parameter predictor, and if only one quantization parameter is available, the available quantization parameter is set as the quantization parameter predictor.

[Second embodiment]

The current coding unit may not have a left coding unit, an upper coding unit, and an upper left coding unit. On the other hand, the current coding unit may have a pre-coding unit in the coding order. Thus, the quantization parameters of the adjacent coding unit and the pre-coding unit immediately adjacent to the current coding unit can be used to generate a quantization parameter predictor. The quantization parameters are retrieved in the following order: 1) quantization parameters of the left adjacent coding unit; 2) quantization parameters of the upper adjacent coding unit; 3) quantization parameters of the upper left adjacent coding unit; and 4) quantization parameters of the pre-coding unit.

Alternatively, the quantization parameters are retrieved in the following order: 1) quantization parameters of the left adjacent coding unit; 2) quantization parameters of the upper adjacent coding unit; and 3) quantization parameters of the pre-coding unit.

When two or more quantization parameters are available, the average of the first two quantization parameters is set as a quantization parameter predictor, and when only one quantization parameter is available, the available quantization parameter is set as a quantization parameter predictor. For example, if the quantization parameters of the left and upper coding units are available, the average of the left and upper quantization parameters is set as the quantization parameter predictor. If only one of the quantization parameters of the left and upper coding units is available, the average of the available quantization parameters and the quantization parameters of the pre-coding unit is set as the quantization parameter predictor. If the quantization parameters of the left and upper coding units are not available, the quantization parameter of the pre-coding unit is set as the quantization parameter predictor. The average is rounded off.

The inverse quantization unit 2030 adds the differential quantization parameter to the quantization parameter predictor A quantization parameter of the current coding unit is generated. If the differential quantization parameter of the current coding unit is not transmitted from the coding side, the differential quantization parameter is set to zero. A quantization parameter is generated for each quantization unit.

The inverse quantization unit 2030 inversely quantizes the quantized block.

The inverse transform unit 2040 inverse transforms the inverse quantized block to generate a residual block. The inverse transform matrix type is determined based on the prediction mode (intra-frame prediction mode and inter-frame prediction mode) and the size of the transform unit.

Adding unit 2090 adds the residual block to the prediction block to produce a reconstructed sample.

The intra-frame prediction unit 2050 restores the intra-frame prediction mode of the current prediction unit based on the intra-frame prediction information received from the entropy decoding unit 2010, and generates a prediction block according to the intra-frame prediction mode.

Inter-frame prediction unit 2060 recovers one or more reference picture indices and one or more motion vectors based on inter-frame prediction information received from entropy decoding unit 2010, and utilizes the one or more reference pictures with one or more The motion vector produces a prediction block.

The operation of the post-processing unit 2070 is the same as the operation of the post-processing unit 1100 of FIG.

The picture storage unit 2080 stores a picture that is post-processed by the post-processing unit 2070.

Figure 4 is a flow chart of the deblocking filtering process in accordance with the present invention.

The post-processing unit 1100 of the dynamic picture encoding device 1000 shown in FIG. 2 and the post-processing unit 2070 of the dynamic picture decoding device 2000 shown in FIG. 3 are executed. Deblocking filtering process.

When it is determined that deblocking filtering is performed on the slice map, the deblocking filtering process is applied to the slice map. The dynamic picture decoding device uses the flag diable_deblocking_filter_flag received from the bit stream to determine whether or not to perform deblocking filtering on each slice.

Deblocking filtering is performed on each coding unit. First, the filtering of the vertical edges is performed from the edge on the left-hand side of the coding unit to the right-hand side of the coding unit. The filtering of the horizontal edges is then carried out from the top edge of the coding unit towards the bottom of the coding unit.

Deblocking filtering is performed only at the edge of the prediction unit and the edge of the transform unit. If the width or height of the prediction unit or transform unit is less than 8 sample lengths, deblocking filtering is performed only at the edges of the 8x8 sample grid.

The boundary strength of each of the 4 sample edges located in the 8x8 sample grid is determined (S110).

Figure 5 is a schematic illustration of a method of determining boundary strength in accordance with the present invention.

As shown in Fig. 5, the boundary strength of each edge is determined in one step in the coding unit. The boundary strength is determined only at the 4 sample edges located in the 8x8 sample grid. The present invention removes the boundary strength on the edges of 4x4 blocks that are not part of the 8x8 sample grid. The present invention also eliminates the operation of creating boundary strength at the edge of 8 samples. Alternatively, in accordance with the present invention, each of the 4 sample edges has its own boundary strength to align with the deblocking decision at the 4 sample edges.

Thus, the computational complexity required to determine the boundary strength in accordance with the present invention is reduced by 50% or more when compared to HEVC being developed. Moreover, the present invention reduces the memory capacity and bandwidth required to determine the boundary strength by 50% or more. Therefore, without deterioration In the case of quality, the present invention reduces the complexity of hardware and software.

Figure 6 is a schematic illustration of a 4-sample edge in accordance with the present invention. As shown in Fig. 6, the 4-sample edge is located between the P block containing the sample p0 and the Q block containing the sample q0. The sample p0 corresponds to one of the samples p0 ₀ to p0 ₃ , and the sample q0 corresponds to one of the samples q0 ₀ to q0 ₃ . Blocks P and Q are prediction blocks or transform blocks.

The boundary strength is determined as follows. The boundary strength is determined at each of the 4 sample edges.

If the prediction unit containing the sample p0 or the prediction unit containing the sample q0 is intra-frame coding, the boundary strength of the 4-sample edge is set equal to 2. The 4 sample edge is the prediction unit edge. That is, if block P and block Q are inter-frame coding, the boundary strength is set equal to 0 or 1.

Set the boundary strength to equal 1 if one or more of the following conditions are met.

1) 4 sample edges are transform unit edges, transform units containing sample p0 or transform units containing q0 contain one or more non-zero transform coefficients.

2) The 4 sample edge is the prediction unit edge, the prediction unit including the sample p0 and the prediction unit including the sample q0 are inter-frame coding, and the prediction unit including the sample p0 has a different reference picture or different from the prediction unit including the sample q0. The number of motion vectors.

3) The prediction unit including the sample p0 and the prediction unit including the sample q0 are inter-frame coding, and the prediction unit including the sample p0 and the prediction unit including the sample q0 have one motion vector, and between the horizontal and vertical components of the motion vector The absolute difference is greater than or equal to a predetermined value (eg, 1 sample). This edge is not part of the horizontal boundary of the LCU.

4) The prediction unit including the sample p0 and the prediction unit including the sample q0 are inter-frame coding, and the prediction unit including the sample p0 and the prediction unit including the sample q0 have two motion vectors, and the prediction unit including the sample p0 and the inclusion sample q0 The prediction unit has at least one identical reference picture, and the absolute difference between the horizontal and vertical components of the two motion vectors corresponding to the same reference picture is greater than or equal to a predetermined value. This edge is not part of the horizontal boundary of the LCU.

As described above, if the 4 sample edges are not on the 8x8 sample grid, the boundary strength is set equal to zero.

On the other hand, when the edge is a horizontal edge of the LCU, and the prediction unit containing the sample p0 is located above the horizontal edge of the LCU, the motion of the prediction unit containing the sample p0 is based on the size and position of the prediction unit containing the sample p0. The information may be replaced with motion information of the left or right adjacent prediction unit including the prediction unit of the sample p0.

Fig. 7 is a schematic diagram showing an exemplary configuration of a prediction unit according to the present invention.

Deblocking filtering is used for the edges of the 8x8 sample grid. Therefore, the encoding device and the decoding device should store the motion information of the higher prediction unit and the upper right prediction unit of the current LCU. The width of each prediction unit and the motion information of each prediction unit are stored to correctly restore motion information. In order to reduce the amount of information stored, each fixed width motion information may be stored to clear the width information to be stored. According to the present invention, the fixed width is set to a plurality of minimum allowable widths (for example, 2 times the minimum allowable width) to reduce the capacity of the line buffer. The fixed width can be set to 8 sample lengths.

Figure 8 is a sports capital to be stored in an online buffer according to the present invention. Schematic diagram of the news.

The upper part of Fig. 8 shows the size and motion information of the upper prediction unit and the upper right prediction unit of the current LCU. The lower part of Fig. 8 shows the motion information to be stored in the line buffer.

As shown in Fig. 8, if the width of a prediction unit is 8 samples, the motion information C is stored at the original width. If the width of the prediction unit is greater than 8, the same motion information H is stored for each fixed width of the length of 8 samples.

However, if the width of the prediction unit is 4 samples long, the motion information to be stored can be replaced based on the position of the prediction unit. For example, the motion information A' is stored for the prediction unit having the motion information A and the prediction unit having the motion information B. The motion information D' is stored for the prediction unit having the motion information D and the prediction unit having the motion information E, and the motion information F' is stored for the prediction unit having the motion information F and the prediction unit having the motion information G.

The motion information A', D', and F' may be set as the motion information of the left prediction unit of the two prediction units.

Alternatively, the motion information A', D', and F' may be set as motion information of the prediction unit adjacent to the vertical line of the 16x16 sample grid of the two prediction units. That is to say, the exercise information A' can be set as the exercise information A, the exercise information D' can be set as the exercise information D, and the exercise information F' can be set as the exercise information G.

Next, it is determined whether deblocking filtering is performed at the edge of 4 samples (S120).

Figure 9 is a schematic illustration of the position of a sample for determining whether a block edge is filtered, in accordance with the present invention. As shown in Figure 5, the implementation is performed at the 4-sample edge of the 8x8 block edge. set.

For each edge, deblocking filtering is performed if the following two conditions are satisfied.

1) bS>0

2) d<β

bS represents the boundary strength. The value of the variable β is determined based on the boundary quantization parameter QP _B .

The variable d is defined as follows.

In the first embodiment, for the region 1, d = d _p0 + d _q0 + d _p3 + d _q3 , and for the region 2, d = d _p4 + d _q4 + d _p7 + d _q7 , d _pk = | p2 _k - 2. P1 _k +p0 _k | and d _qk =| q2 _k -2. Q1 _k +q0 _k |.

In the second embodiment, for the region 1, d = d _p0 + d _q0 + d _p2 + d _q2 , and for the region 2, d = d _p4 + d _q4 + d _p6 + d _q6 .

In the third embodiment, for the region 1, d = d _p1 + d _q1 + d _p2 + d _q2 , and for the region 2, d = d _p5 + d _q5 + d _p6 + d _q6 .

Next, if it is determined that deblocking filtering is used for the 4-sample edge, one deblocking filter is selected among the strong filter and the weak filter. However, if it is determined that deblocking filtering is not applied to the 4-sample edge, the deblocking filtering process ends for this edge. As shown in Figure 9, a filter is selected for each of the 4 sample edges.

Select a strong filter for Area 1 if the following conditions are met.

1)d<(β>>2)

2)| p3 _i -p0 _i |+| q3 _i -q0 _i |<(β>>3), where i=0,3

3)| p0 _i -q0 _i |<(5*t _c +1)>>1, where i=0,3 or

1) d _i <(β>>1), where i=0,3

2)| p3 _i -p0 _i |+| q3 _i -q0 _i |<(β>>3), where i=0,3

3)| p0 _i -q0 _i |<(5*t _c +1)>>1, where i=0,3

Otherwise, choose a weak filter. The value of the variable tc is determined based on the boundary quantization parameter QP _B .

Select a strong filter for Region 2 if the following conditions are met.

1)d<(β>>2)

2)| p3 _i -p0i |+| q3i-q0i |<(β>>3), where i=4,7

3)| p0i-q0i |<(5*tc+1)>>1, where i=4,7

or

1) d _i <(β>>1), where i=4,7

2)| p3 _i -p0 _i |+| q3 _i -q0 _i |<(β>>3), where i=4,7

3)| p0 _i -q0 _i |<(5*t _c +1)>>1, where i=4,7

Otherwise, choose a weak filter.

In another embodiment, for region 1, i = 0, 3 is substituted for i = 0, 3, and for region 2, i = 5, 7 is substituted for i = 4, 7.

In another embodiment, i = 1, 2 is replaced by i = 1, 3 for region 1, and i = 4, 7 is replaced by i = 5, 6 for region 2.

Next, if the block filter is selected to be understood, the edge is filtered by the deblocking filter (S140).

The strong filter is as follows.

p ₀ '=(p ₂ +2*p ₁ +2*p ₀ +2*q ₀ +q ₁ +4)>>3)

p ₁ '=(p ₂ +p ₁ +p ₀ +q ₀ +2)>>2

p ₂ '=(2*p ₃ +3*p ₂ +p ₁ +p ₀ +q ₀ +4)>>3

q ₀ '=(p ₁ +2*p ₀ +2*q ₀ +2*q ₁ +q ₂ +4)>>3

q ₁ '=(p ₀ +q ₀ +q ₁ +q ₂ +2)>>2

q ₂ '=(p ₀ +q ₀ +q ₁ +3*q ₂ +2*q ₃ +4)>>3

The weak filter is as follows.

△=Clip3(-t _C , t _C , △)

p ₀ '=Clip1(p ₀ +△)

q ₀ '=Clip1(q ₀ -△)

Δp=Clip3(-(t _C >>1), t _C >>1,(((p ₂ +p ₀ +1)>>1)-p ₁ +△)>>1)

p ₁ '=Clip1(p ₁ +Δp)

Δq=Clip3(-(t _C >>1), t _C >>1,(((q ₂ +q ₀ +1)>>1)-q ₁ -△)>>1)

q ₁ '=Clip1(q ₁ +Δq)

The variables β and t _{C are} determined by the boundary quantization parameter QP _B and monotonically increase as the boundary quantization parameter QP _B increases. The relationship between the parameters β and t _C and the quantization parameter are defined as a table.

The boundary quantization parameter QPB is an average of the quantization parameter QP _P of the P block including the sample p0 and the quantization parameter QP _Q of the Q block including the sample q0. This average is a rounded value. If at least one of the P block and the Q block is intra-frame coded, when QP _{B is} increased by 1, the parameter t _{C is} increased by 0, 1, or 2.

Although the present invention has been shown and described with reference to certain exemplary embodiments thereof, it will be understood by those skilled in the art Spirit and scope.

Claims (10)

A method for removing blocky artifacts includes: determining a boundary strength of each of the 4 sample edges located at an edge of the transform unit or the prediction unit and located on the 8×8 sample grid; if the boundary strength of the 4 sample edges If not equal to 0, it is determined whether deblocking filtering is performed at the 4 sample edge; if the deblocking filtering is performed at the 4 sample edge, selecting a deblocking filter suitable for the 4 sample edge; using the selected filter The 4 sample edges are filtered; and the boundary strength is not determined for 4 sample edges that are not located on the 8x8 sample grid. A method of clearing block human noise according to claim 1 of the patent application, wherein the boundary strength is set equal to 0 if the 4 sample edge is not a transform unit edge and the 4 sample edge is not a prediction unit edge. A method of clearing block human noise according to item 1 of the patent application, wherein the boundary strength is set equal to 0, 1, or 2. According to the method of the first aspect of the patent application, the method of clearing block human noise is performed, wherein after filtering the vertical 4 sample edges, the horizontal 4 sample edges are filtered. The method for clearing block-by-human noise according to item 4 of the patent application scope, wherein if the prediction unit including the same p0 or the prediction unit including the same q0 is intra-frame coding, the boundary strength is set equal to 2 . The method for clearing blocky artificial noise according to item 5 of the patent application scope, wherein if the prediction unit including the sample p0 and the prediction unit including the sample q0 are inter-frame coding, the boundary strength is set equal to 1 Or 0. A method of clearing blocky artifacts according to claim 1 of the scope of the patent application, wherein boundary quantization parameters are used to determine whether deblocking filtering is performed at the 4-sample edge. According to the method of claim 7, the method for clearing block human noise is set, wherein the boundary quantization parameter is set to an average value of a quantization parameter including a block of the same p0 and a quantization parameter including a block of the same q0. The method for clearing block human noise according to item 8 of the patent application scope, wherein the average value is a rounded value. A method of clearing block human noise according to item 8 of the patent application scope, wherein a deblocking filter is selected using boundary quantization parameters.